Commit | Line | Data |
---|---|---|
60c778b2 | 1 | /* SCTP kernel implementation |
1da177e4 LT |
2 | * Copyright (c) 1999-2000 Cisco, Inc. |
3 | * Copyright (c) 1999-2001 Motorola, Inc. | |
4 | * Copyright (c) 2001-2003 International Business Machines Corp. | |
5 | * Copyright (c) 2001 Intel Corp. | |
6 | * Copyright (c) 2001 La Monte H.P. Yarroll | |
7 | * | |
60c778b2 | 8 | * This file is part of the SCTP kernel implementation |
1da177e4 LT |
9 | * |
10 | * This module provides the abstraction for an SCTP tranport representing | |
11 | * a remote transport address. For local transport addresses, we just use | |
12 | * union sctp_addr. | |
13 | * | |
60c778b2 | 14 | * This SCTP implementation is free software; |
1da177e4 LT |
15 | * you can redistribute it and/or modify it under the terms of |
16 | * the GNU General Public License as published by | |
17 | * the Free Software Foundation; either version 2, or (at your option) | |
18 | * any later version. | |
19 | * | |
60c778b2 | 20 | * This SCTP implementation is distributed in the hope that it |
1da177e4 LT |
21 | * will be useful, but WITHOUT ANY WARRANTY; without even the implied |
22 | * ************************ | |
23 | * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. | |
24 | * See the GNU General Public License for more details. | |
25 | * | |
26 | * You should have received a copy of the GNU General Public License | |
27 | * along with GNU CC; see the file COPYING. If not, write to | |
28 | * the Free Software Foundation, 59 Temple Place - Suite 330, | |
29 | * Boston, MA 02111-1307, USA. | |
30 | * | |
31 | * Please send any bug reports or fixes you make to the | |
32 | * email address(es): | |
33 | * lksctp developers <lksctp-developers@lists.sourceforge.net> | |
34 | * | |
35 | * Or submit a bug report through the following website: | |
36 | * http://www.sf.net/projects/lksctp | |
37 | * | |
38 | * Written or modified by: | |
39 | * La Monte H.P. Yarroll <piggy@acm.org> | |
40 | * Karl Knutson <karl@athena.chicago.il.us> | |
41 | * Jon Grimm <jgrimm@us.ibm.com> | |
42 | * Xingang Guo <xingang.guo@intel.com> | |
43 | * Hui Huang <hui.huang@nokia.com> | |
44 | * Sridhar Samudrala <sri@us.ibm.com> | |
45 | * Ardelle Fan <ardelle.fan@intel.com> | |
46 | * | |
47 | * Any bugs reported given to us we will try to fix... any fixes shared will | |
48 | * be incorporated into the next SCTP release. | |
49 | */ | |
50 | ||
145ce502 JP |
51 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
52 | ||
5a0e3ad6 | 53 | #include <linux/slab.h> |
1da177e4 | 54 | #include <linux/types.h> |
ad8fec17 | 55 | #include <linux/random.h> |
1da177e4 LT |
56 | #include <net/sctp/sctp.h> |
57 | #include <net/sctp/sm.h> | |
58 | ||
59 | /* 1st Level Abstractions. */ | |
60 | ||
61 | /* Initialize a new transport from provided memory. */ | |
62 | static struct sctp_transport *sctp_transport_init(struct sctp_transport *peer, | |
63 | const union sctp_addr *addr, | |
dd0fc66f | 64 | gfp_t gfp) |
1da177e4 LT |
65 | { |
66 | /* Copy in the address. */ | |
67 | peer->ipaddr = *addr; | |
68 | peer->af_specific = sctp_get_af_specific(addr->sa.sa_family); | |
1da177e4 LT |
69 | memset(&peer->saddr, 0, sizeof(union sctp_addr)); |
70 | ||
4244854d NH |
71 | peer->sack_generation = 0; |
72 | ||
1da177e4 LT |
73 | /* From 6.3.1 RTO Calculation: |
74 | * | |
75 | * C1) Until an RTT measurement has been made for a packet sent to the | |
76 | * given destination transport address, set RTO to the protocol | |
77 | * parameter 'RTO.Initial'. | |
78 | */ | |
5fdd4bae | 79 | peer->rto = msecs_to_jiffies(sctp_rto_initial); |
1da177e4 LT |
80 | |
81 | peer->last_time_heard = jiffies; | |
1da177e4 LT |
82 | peer->last_time_ecne_reduced = jiffies; |
83 | ||
52ccb8e9 FF |
84 | peer->param_flags = SPP_HB_DISABLE | |
85 | SPP_PMTUD_ENABLE | | |
86 | SPP_SACKDELAY_ENABLE; | |
1da177e4 LT |
87 | |
88 | /* Initialize the default path max_retrans. */ | |
52ccb8e9 | 89 | peer->pathmaxrxt = sctp_max_retrans_path; |
1da177e4 LT |
90 | |
91 | INIT_LIST_HEAD(&peer->transmitted); | |
92 | INIT_LIST_HEAD(&peer->send_ready); | |
93 | INIT_LIST_HEAD(&peer->transports); | |
94 | ||
b24b8a24 PE |
95 | setup_timer(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event, |
96 | (unsigned long)peer); | |
97 | setup_timer(&peer->hb_timer, sctp_generate_heartbeat_event, | |
98 | (unsigned long)peer); | |
50b5d6ad VY |
99 | setup_timer(&peer->proto_unreach_timer, |
100 | sctp_generate_proto_unreach_event, (unsigned long)peer); | |
1da177e4 | 101 | |
ad8fec17 SS |
102 | /* Initialize the 64-bit random nonce sent with heartbeat. */ |
103 | get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce)); | |
104 | ||
1da177e4 | 105 | atomic_set(&peer->refcnt, 1); |
1da177e4 LT |
106 | |
107 | return peer; | |
108 | } | |
109 | ||
110 | /* Allocate and initialize a new transport. */ | |
3182cd84 | 111 | struct sctp_transport *sctp_transport_new(const union sctp_addr *addr, |
dd0fc66f | 112 | gfp_t gfp) |
1da177e4 | 113 | { |
d808ad9a | 114 | struct sctp_transport *transport; |
1da177e4 | 115 | |
d808ad9a | 116 | transport = t_new(struct sctp_transport, gfp); |
1da177e4 LT |
117 | if (!transport) |
118 | goto fail; | |
119 | ||
120 | if (!sctp_transport_init(transport, addr, gfp)) | |
121 | goto fail_init; | |
122 | ||
123 | transport->malloced = 1; | |
124 | SCTP_DBG_OBJCNT_INC(transport); | |
125 | ||
126 | return transport; | |
127 | ||
128 | fail_init: | |
129 | kfree(transport); | |
130 | ||
131 | fail: | |
132 | return NULL; | |
133 | } | |
134 | ||
135 | /* This transport is no longer needed. Free up if possible, or | |
136 | * delay until it last reference count. | |
137 | */ | |
138 | void sctp_transport_free(struct sctp_transport *transport) | |
139 | { | |
140 | transport->dead = 1; | |
141 | ||
142 | /* Try to delete the heartbeat timer. */ | |
143 | if (del_timer(&transport->hb_timer)) | |
144 | sctp_transport_put(transport); | |
145 | ||
146 | /* Delete the T3_rtx timer if it's active. | |
147 | * There is no point in not doing this now and letting | |
148 | * structure hang around in memory since we know | |
149 | * the tranport is going away. | |
150 | */ | |
151 | if (timer_pending(&transport->T3_rtx_timer) && | |
152 | del_timer(&transport->T3_rtx_timer)) | |
153 | sctp_transport_put(transport); | |
154 | ||
55fa0cfd WY |
155 | /* Delete the ICMP proto unreachable timer if it's active. */ |
156 | if (timer_pending(&transport->proto_unreach_timer) && | |
157 | del_timer(&transport->proto_unreach_timer)) | |
158 | sctp_association_put(transport->asoc); | |
1da177e4 LT |
159 | |
160 | sctp_transport_put(transport); | |
161 | } | |
162 | ||
163 | /* Destroy the transport data structure. | |
164 | * Assumes there are no more users of this structure. | |
165 | */ | |
166 | static void sctp_transport_destroy(struct sctp_transport *transport) | |
167 | { | |
168 | SCTP_ASSERT(transport->dead, "Transport is not dead", return); | |
169 | ||
170 | if (transport->asoc) | |
171 | sctp_association_put(transport->asoc); | |
172 | ||
d808ad9a | 173 | sctp_packet_free(&transport->packet); |
1da177e4 LT |
174 | |
175 | dst_release(transport->dst); | |
176 | kfree(transport); | |
177 | SCTP_DBG_OBJCNT_DEC(transport); | |
178 | } | |
179 | ||
180 | /* Start T3_rtx timer if it is not already running and update the heartbeat | |
181 | * timer. This routine is called every time a DATA chunk is sent. | |
182 | */ | |
d9efc223 | 183 | void sctp_transport_reset_timers(struct sctp_transport *transport) |
1da177e4 LT |
184 | { |
185 | /* RFC 2960 6.3.2 Retransmission Timer Rules | |
186 | * | |
187 | * R1) Every time a DATA chunk is sent to any address(including a | |
188 | * retransmission), if the T3-rtx timer of that address is not running | |
189 | * start it running so that it will expire after the RTO of that | |
190 | * address. | |
191 | */ | |
192 | ||
d9efc223 | 193 | if (!timer_pending(&transport->T3_rtx_timer)) |
1da177e4 LT |
194 | if (!mod_timer(&transport->T3_rtx_timer, |
195 | jiffies + transport->rto)) | |
196 | sctp_transport_hold(transport); | |
197 | ||
198 | /* When a data chunk is sent, reset the heartbeat interval. */ | |
199 | if (!mod_timer(&transport->hb_timer, | |
200 | sctp_transport_timeout(transport))) | |
201 | sctp_transport_hold(transport); | |
202 | } | |
203 | ||
204 | /* This transport has been assigned to an association. | |
205 | * Initialize fields from the association or from the sock itself. | |
206 | * Register the reference count in the association. | |
207 | */ | |
208 | void sctp_transport_set_owner(struct sctp_transport *transport, | |
209 | struct sctp_association *asoc) | |
210 | { | |
211 | transport->asoc = asoc; | |
212 | sctp_association_hold(asoc); | |
213 | } | |
214 | ||
215 | /* Initialize the pmtu of a transport. */ | |
9914ae3c | 216 | void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk) |
1da177e4 | 217 | { |
da0420be VY |
218 | /* If we don't have a fresh route, look one up */ |
219 | if (!transport->dst || transport->dst->obsolete > 1) { | |
220 | dst_release(transport->dst); | |
221 | transport->af_specific->get_dst(transport, &transport->saddr, | |
8663c938 | 222 | &transport->fl, sk); |
da0420be | 223 | } |
1da177e4 | 224 | |
da0420be VY |
225 | if (transport->dst) { |
226 | transport->pathmtu = dst_mtu(transport->dst); | |
1da177e4 | 227 | } else |
52ccb8e9 | 228 | transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT; |
1da177e4 LT |
229 | } |
230 | ||
c910b47e VY |
231 | void sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu) |
232 | { | |
233 | struct dst_entry *dst; | |
234 | ||
235 | if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) { | |
145ce502 JP |
236 | pr_warn("%s: Reported pmtu %d too low, using default minimum of %d\n", |
237 | __func__, pmtu, | |
238 | SCTP_DEFAULT_MINSEGMENT); | |
c910b47e VY |
239 | /* Use default minimum segment size and disable |
240 | * pmtu discovery on this transport. | |
241 | */ | |
242 | t->pathmtu = SCTP_DEFAULT_MINSEGMENT; | |
c910b47e VY |
243 | } else { |
244 | t->pathmtu = pmtu; | |
245 | } | |
246 | ||
247 | dst = sctp_transport_dst_check(t); | |
248 | if (dst) | |
249 | dst->ops->update_pmtu(dst, pmtu); | |
250 | } | |
251 | ||
1da177e4 LT |
252 | /* Caches the dst entry and source address for a transport's destination |
253 | * address. | |
254 | */ | |
255 | void sctp_transport_route(struct sctp_transport *transport, | |
256 | union sctp_addr *saddr, struct sctp_sock *opt) | |
257 | { | |
258 | struct sctp_association *asoc = transport->asoc; | |
259 | struct sctp_af *af = transport->af_specific; | |
1da177e4 | 260 | |
8663c938 | 261 | af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt)); |
1da177e4 LT |
262 | |
263 | if (saddr) | |
264 | memcpy(&transport->saddr, saddr, sizeof(union sctp_addr)); | |
265 | else | |
8663c938 | 266 | af->get_saddr(opt, transport, &transport->fl); |
1da177e4 | 267 | |
52ccb8e9 FF |
268 | if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) { |
269 | return; | |
270 | } | |
da0420be VY |
271 | if (transport->dst) { |
272 | transport->pathmtu = dst_mtu(transport->dst); | |
1da177e4 LT |
273 | |
274 | /* Initialize sk->sk_rcv_saddr, if the transport is the | |
275 | * association's active path for getsockname(). | |
d808ad9a | 276 | */ |
a78102e7 VY |
277 | if (asoc && (!asoc->peer.primary_path || |
278 | (transport == asoc->peer.active_path))) | |
bf031fff NH |
279 | opt->pf->af->to_sk_saddr(&transport->saddr, |
280 | asoc->base.sk); | |
1da177e4 | 281 | } else |
52ccb8e9 | 282 | transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT; |
1da177e4 LT |
283 | } |
284 | ||
285 | /* Hold a reference to a transport. */ | |
286 | void sctp_transport_hold(struct sctp_transport *transport) | |
287 | { | |
288 | atomic_inc(&transport->refcnt); | |
289 | } | |
290 | ||
291 | /* Release a reference to a transport and clean up | |
292 | * if there are no more references. | |
293 | */ | |
294 | void sctp_transport_put(struct sctp_transport *transport) | |
295 | { | |
296 | if (atomic_dec_and_test(&transport->refcnt)) | |
297 | sctp_transport_destroy(transport); | |
298 | } | |
299 | ||
300 | /* Update transport's RTO based on the newly calculated RTT. */ | |
301 | void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt) | |
302 | { | |
303 | /* Check for valid transport. */ | |
304 | SCTP_ASSERT(tp, "NULL transport", return); | |
305 | ||
306 | /* We should not be doing any RTO updates unless rto_pending is set. */ | |
307 | SCTP_ASSERT(tp->rto_pending, "rto_pending not set", return); | |
308 | ||
309 | if (tp->rttvar || tp->srtt) { | |
310 | /* 6.3.1 C3) When a new RTT measurement R' is made, set | |
311 | * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'| | |
312 | * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R' | |
313 | */ | |
314 | ||
315 | /* Note: The above algorithm has been rewritten to | |
316 | * express rto_beta and rto_alpha as inverse powers | |
317 | * of two. | |
318 | * For example, assuming the default value of RTO.Alpha of | |
319 | * 1/8, rto_alpha would be expressed as 3. | |
320 | */ | |
321 | tp->rttvar = tp->rttvar - (tp->rttvar >> sctp_rto_beta) | |
322 | + ((abs(tp->srtt - rtt)) >> sctp_rto_beta); | |
323 | tp->srtt = tp->srtt - (tp->srtt >> sctp_rto_alpha) | |
324 | + (rtt >> sctp_rto_alpha); | |
325 | } else { | |
326 | /* 6.3.1 C2) When the first RTT measurement R is made, set | |
327 | * SRTT <- R, RTTVAR <- R/2. | |
328 | */ | |
329 | tp->srtt = rtt; | |
330 | tp->rttvar = rtt >> 1; | |
331 | } | |
332 | ||
333 | /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then | |
334 | * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY. | |
335 | */ | |
336 | if (tp->rttvar == 0) | |
337 | tp->rttvar = SCTP_CLOCK_GRANULARITY; | |
338 | ||
339 | /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */ | |
340 | tp->rto = tp->srtt + (tp->rttvar << 2); | |
341 | ||
342 | /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min | |
343 | * seconds then it is rounded up to RTO.Min seconds. | |
344 | */ | |
345 | if (tp->rto < tp->asoc->rto_min) | |
346 | tp->rto = tp->asoc->rto_min; | |
347 | ||
348 | /* 6.3.1 C7) A maximum value may be placed on RTO provided it is | |
349 | * at least RTO.max seconds. | |
350 | */ | |
351 | if (tp->rto > tp->asoc->rto_max) | |
352 | tp->rto = tp->asoc->rto_max; | |
353 | ||
354 | tp->rtt = rtt; | |
355 | ||
356 | /* Reset rto_pending so that a new RTT measurement is started when a | |
357 | * new data chunk is sent. | |
358 | */ | |
359 | tp->rto_pending = 0; | |
360 | ||
361 | SCTP_DEBUG_PRINTK("%s: transport: %p, rtt: %d, srtt: %d " | |
0dc47877 | 362 | "rttvar: %d, rto: %ld\n", __func__, |
1da177e4 LT |
363 | tp, rtt, tp->srtt, tp->rttvar, tp->rto); |
364 | } | |
365 | ||
366 | /* This routine updates the transport's cwnd and partial_bytes_acked | |
367 | * parameters based on the bytes acked in the received SACK. | |
368 | */ | |
369 | void sctp_transport_raise_cwnd(struct sctp_transport *transport, | |
370 | __u32 sack_ctsn, __u32 bytes_acked) | |
371 | { | |
cf9b4812 | 372 | struct sctp_association *asoc = transport->asoc; |
1da177e4 LT |
373 | __u32 cwnd, ssthresh, flight_size, pba, pmtu; |
374 | ||
375 | cwnd = transport->cwnd; | |
376 | flight_size = transport->flight_size; | |
377 | ||
a6465234 | 378 | /* See if we need to exit Fast Recovery first */ |
cf9b4812 VY |
379 | if (asoc->fast_recovery && |
380 | TSN_lte(asoc->fast_recovery_exit, sack_ctsn)) | |
381 | asoc->fast_recovery = 0; | |
a6465234 | 382 | |
1da177e4 | 383 | /* The appropriate cwnd increase algorithm is performed if, and only |
a6465234 | 384 | * if the cumulative TSN whould advanced and the congestion window is |
1da177e4 LT |
385 | * being fully utilized. |
386 | */ | |
a6465234 | 387 | if (TSN_lte(sack_ctsn, transport->asoc->ctsn_ack_point) || |
1da177e4 LT |
388 | (flight_size < cwnd)) |
389 | return; | |
390 | ||
391 | ssthresh = transport->ssthresh; | |
392 | pba = transport->partial_bytes_acked; | |
52ccb8e9 | 393 | pmtu = transport->asoc->pathmtu; |
1da177e4 LT |
394 | |
395 | if (cwnd <= ssthresh) { | |
a6465234 VY |
396 | /* RFC 4960 7.2.1 |
397 | * o When cwnd is less than or equal to ssthresh, an SCTP | |
398 | * endpoint MUST use the slow-start algorithm to increase | |
399 | * cwnd only if the current congestion window is being fully | |
400 | * utilized, an incoming SACK advances the Cumulative TSN | |
401 | * Ack Point, and the data sender is not in Fast Recovery. | |
402 | * Only when these three conditions are met can the cwnd be | |
403 | * increased; otherwise, the cwnd MUST not be increased. | |
404 | * If these conditions are met, then cwnd MUST be increased | |
405 | * by, at most, the lesser of 1) the total size of the | |
406 | * previously outstanding DATA chunk(s) acknowledged, and | |
407 | * 2) the destination's path MTU. This upper bound protects | |
408 | * against the ACK-Splitting attack outlined in [SAVAGE99]. | |
1da177e4 | 409 | */ |
cf9b4812 | 410 | if (asoc->fast_recovery) |
a6465234 VY |
411 | return; |
412 | ||
1da177e4 LT |
413 | if (bytes_acked > pmtu) |
414 | cwnd += pmtu; | |
415 | else | |
416 | cwnd += bytes_acked; | |
417 | SCTP_DEBUG_PRINTK("%s: SLOW START: transport: %p, " | |
418 | "bytes_acked: %d, cwnd: %d, ssthresh: %d, " | |
419 | "flight_size: %d, pba: %d\n", | |
0dc47877 | 420 | __func__, |
1da177e4 LT |
421 | transport, bytes_acked, cwnd, |
422 | ssthresh, flight_size, pba); | |
423 | } else { | |
424 | /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh, | |
425 | * upon each SACK arrival that advances the Cumulative TSN Ack | |
426 | * Point, increase partial_bytes_acked by the total number of | |
427 | * bytes of all new chunks acknowledged in that SACK including | |
428 | * chunks acknowledged by the new Cumulative TSN Ack and by | |
429 | * Gap Ack Blocks. | |
430 | * | |
431 | * When partial_bytes_acked is equal to or greater than cwnd | |
432 | * and before the arrival of the SACK the sender had cwnd or | |
433 | * more bytes of data outstanding (i.e., before arrival of the | |
434 | * SACK, flightsize was greater than or equal to cwnd), | |
435 | * increase cwnd by MTU, and reset partial_bytes_acked to | |
436 | * (partial_bytes_acked - cwnd). | |
437 | */ | |
438 | pba += bytes_acked; | |
439 | if (pba >= cwnd) { | |
440 | cwnd += pmtu; | |
441 | pba = ((cwnd < pba) ? (pba - cwnd) : 0); | |
442 | } | |
443 | SCTP_DEBUG_PRINTK("%s: CONGESTION AVOIDANCE: " | |
444 | "transport: %p, bytes_acked: %d, cwnd: %d, " | |
445 | "ssthresh: %d, flight_size: %d, pba: %d\n", | |
0dc47877 | 446 | __func__, |
1da177e4 LT |
447 | transport, bytes_acked, cwnd, |
448 | ssthresh, flight_size, pba); | |
449 | } | |
450 | ||
451 | transport->cwnd = cwnd; | |
452 | transport->partial_bytes_acked = pba; | |
453 | } | |
454 | ||
455 | /* This routine is used to lower the transport's cwnd when congestion is | |
456 | * detected. | |
457 | */ | |
458 | void sctp_transport_lower_cwnd(struct sctp_transport *transport, | |
459 | sctp_lower_cwnd_t reason) | |
460 | { | |
cf9b4812 VY |
461 | struct sctp_association *asoc = transport->asoc; |
462 | ||
1da177e4 LT |
463 | switch (reason) { |
464 | case SCTP_LOWER_CWND_T3_RTX: | |
465 | /* RFC 2960 Section 7.2.3, sctpimpguide | |
466 | * When the T3-rtx timer expires on an address, SCTP should | |
467 | * perform slow start by: | |
468 | * ssthresh = max(cwnd/2, 4*MTU) | |
469 | * cwnd = 1*MTU | |
470 | * partial_bytes_acked = 0 | |
471 | */ | |
472 | transport->ssthresh = max(transport->cwnd/2, | |
cf9b4812 VY |
473 | 4*asoc->pathmtu); |
474 | transport->cwnd = asoc->pathmtu; | |
33ce8281 | 475 | |
cf9b4812 VY |
476 | /* T3-rtx also clears fast recovery */ |
477 | asoc->fast_recovery = 0; | |
1da177e4 LT |
478 | break; |
479 | ||
480 | case SCTP_LOWER_CWND_FAST_RTX: | |
481 | /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the | |
482 | * destination address(es) to which the missing DATA chunks | |
483 | * were last sent, according to the formula described in | |
484 | * Section 7.2.3. | |
d808ad9a YH |
485 | * |
486 | * RFC 2960 7.2.3, sctpimpguide Upon detection of packet | |
1da177e4 LT |
487 | * losses from SACK (see Section 7.2.4), An endpoint |
488 | * should do the following: | |
489 | * ssthresh = max(cwnd/2, 4*MTU) | |
490 | * cwnd = ssthresh | |
491 | * partial_bytes_acked = 0 | |
492 | */ | |
cf9b4812 | 493 | if (asoc->fast_recovery) |
a6465234 VY |
494 | return; |
495 | ||
496 | /* Mark Fast recovery */ | |
cf9b4812 VY |
497 | asoc->fast_recovery = 1; |
498 | asoc->fast_recovery_exit = asoc->next_tsn - 1; | |
a6465234 | 499 | |
1da177e4 | 500 | transport->ssthresh = max(transport->cwnd/2, |
cf9b4812 | 501 | 4*asoc->pathmtu); |
1da177e4 LT |
502 | transport->cwnd = transport->ssthresh; |
503 | break; | |
504 | ||
505 | case SCTP_LOWER_CWND_ECNE: | |
506 | /* RFC 2481 Section 6.1.2. | |
507 | * If the sender receives an ECN-Echo ACK packet | |
508 | * then the sender knows that congestion was encountered in the | |
509 | * network on the path from the sender to the receiver. The | |
510 | * indication of congestion should be treated just as a | |
511 | * congestion loss in non-ECN Capable TCP. That is, the TCP | |
512 | * source halves the congestion window "cwnd" and reduces the | |
513 | * slow start threshold "ssthresh". | |
514 | * A critical condition is that TCP does not react to | |
515 | * congestion indications more than once every window of | |
516 | * data (or more loosely more than once every round-trip time). | |
517 | */ | |
f61f6f82 WY |
518 | if (time_after(jiffies, transport->last_time_ecne_reduced + |
519 | transport->rtt)) { | |
1da177e4 | 520 | transport->ssthresh = max(transport->cwnd/2, |
cf9b4812 | 521 | 4*asoc->pathmtu); |
1da177e4 LT |
522 | transport->cwnd = transport->ssthresh; |
523 | transport->last_time_ecne_reduced = jiffies; | |
524 | } | |
525 | break; | |
526 | ||
527 | case SCTP_LOWER_CWND_INACTIVE: | |
528 | /* RFC 2960 Section 7.2.1, sctpimpguide | |
529 | * When the endpoint does not transmit data on a given | |
530 | * transport address, the cwnd of the transport address | |
531 | * should be adjusted to max(cwnd/2, 4*MTU) per RTO. | |
532 | * NOTE: Although the draft recommends that this check needs | |
533 | * to be done every RTO interval, we do it every hearbeat | |
534 | * interval. | |
535 | */ | |
245cba7e | 536 | transport->cwnd = max(transport->cwnd/2, |
cf9b4812 | 537 | 4*asoc->pathmtu); |
1da177e4 | 538 | break; |
3ff50b79 | 539 | } |
1da177e4 LT |
540 | |
541 | transport->partial_bytes_acked = 0; | |
542 | SCTP_DEBUG_PRINTK("%s: transport: %p reason: %d cwnd: " | |
0dc47877 | 543 | "%d ssthresh: %d\n", __func__, |
1da177e4 LT |
544 | transport, reason, |
545 | transport->cwnd, transport->ssthresh); | |
546 | } | |
547 | ||
46d5a808 VY |
548 | /* Apply Max.Burst limit to the congestion window: |
549 | * sctpimpguide-05 2.14.2 | |
550 | * D) When the time comes for the sender to | |
551 | * transmit new DATA chunks, the protocol parameter Max.Burst MUST | |
552 | * first be applied to limit how many new DATA chunks may be sent. | |
553 | * The limit is applied by adjusting cwnd as follows: | |
554 | * if ((flightsize+ Max.Burst * MTU) < cwnd) | |
555 | * cwnd = flightsize + Max.Burst * MTU | |
556 | */ | |
557 | ||
558 | void sctp_transport_burst_limited(struct sctp_transport *t) | |
559 | { | |
560 | struct sctp_association *asoc = t->asoc; | |
561 | u32 old_cwnd = t->cwnd; | |
562 | u32 max_burst_bytes; | |
563 | ||
564 | if (t->burst_limited) | |
565 | return; | |
566 | ||
567 | max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu); | |
568 | if (max_burst_bytes < old_cwnd) { | |
569 | t->cwnd = max_burst_bytes; | |
570 | t->burst_limited = old_cwnd; | |
571 | } | |
572 | } | |
573 | ||
574 | /* Restore the old cwnd congestion window, after the burst had it's | |
575 | * desired effect. | |
576 | */ | |
577 | void sctp_transport_burst_reset(struct sctp_transport *t) | |
578 | { | |
579 | if (t->burst_limited) { | |
580 | t->cwnd = t->burst_limited; | |
581 | t->burst_limited = 0; | |
582 | } | |
583 | } | |
584 | ||
1da177e4 LT |
585 | /* What is the next timeout value for this transport? */ |
586 | unsigned long sctp_transport_timeout(struct sctp_transport *t) | |
587 | { | |
588 | unsigned long timeout; | |
ad8fec17 SS |
589 | timeout = t->rto + sctp_jitter(t->rto); |
590 | if (t->state != SCTP_UNCONFIRMED) | |
591 | timeout += t->hbinterval; | |
1da177e4 LT |
592 | timeout += jiffies; |
593 | return timeout; | |
594 | } | |
749bf921 VY |
595 | |
596 | /* Reset transport variables to their initial values */ | |
597 | void sctp_transport_reset(struct sctp_transport *t) | |
598 | { | |
599 | struct sctp_association *asoc = t->asoc; | |
600 | ||
601 | /* RFC 2960 (bis), Section 5.2.4 | |
602 | * All the congestion control parameters (e.g., cwnd, ssthresh) | |
603 | * related to this peer MUST be reset to their initial values | |
604 | * (see Section 6.2.1) | |
605 | */ | |
606 | t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380)); | |
46d5a808 | 607 | t->burst_limited = 0; |
289f4249 | 608 | t->ssthresh = asoc->peer.i.a_rwnd; |
5fdd4bae | 609 | t->rto = asoc->rto_initial; |
749bf921 VY |
610 | t->rtt = 0; |
611 | t->srtt = 0; | |
612 | t->rttvar = 0; | |
613 | ||
614 | /* Reset these additional varibles so that we have a clean | |
615 | * slate. | |
616 | */ | |
617 | t->partial_bytes_acked = 0; | |
618 | t->flight_size = 0; | |
619 | t->error_count = 0; | |
620 | t->rto_pending = 0; | |
faee47cd | 621 | t->hb_sent = 0; |
749bf921 VY |
622 | |
623 | /* Initialize the state information for SFR-CACC */ | |
624 | t->cacc.changeover_active = 0; | |
625 | t->cacc.cycling_changeover = 0; | |
626 | t->cacc.next_tsn_at_change = 0; | |
627 | t->cacc.cacc_saw_newack = 0; | |
628 | } | |
ddc4bbee MH |
629 | |
630 | /* Schedule retransmission on the given transport */ | |
631 | void sctp_transport_immediate_rtx(struct sctp_transport *t) | |
632 | { | |
633 | /* Stop pending T3_rtx_timer */ | |
634 | if (timer_pending(&t->T3_rtx_timer)) { | |
635 | (void)del_timer(&t->T3_rtx_timer); | |
636 | sctp_transport_put(t); | |
637 | } | |
638 | sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX); | |
639 | if (!timer_pending(&t->T3_rtx_timer)) { | |
640 | if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto)) | |
641 | sctp_transport_hold(t); | |
642 | } | |
643 | return; | |
644 | } |